C h a p t e r 1
Given the extremely data-intensive nature
of AI workloads, memory and storage
have become essential components in
modern edge AI applications. In this
context, these technologies are necessary
to enable the processing, movement, and
retention of the vast amounts of data
required for AI workloads.
Memory refers to the temporary, high-
speed storage used by a system to hold
data and instructions that are actively
being processed. In relation to edge AI,
memory acts as the intermediary between
storage and processing units, facilitating
rapid data access and manipulation
essential for AI computations. While
processors handle the actual computations,
they rely on memory to supply the
necessary data quickly and efficiently.
Within this context, dynamic random
access memory (DRAM) is one of the
most important and ubiquitous forms
of memory in AI systems. The speed
and bandwidth of DRAM are particularly
important for AI workloads, as they
directly impact the system's ability to
process large amounts of data quickly.
For this reason, various forms of DRAM
UNLOCKING THE POWER:
MEMORY AND STORAGE IN
AI APPLICATIONS
Peter Müller
VP Product Center Modules, Kontron
Memory and storage solutions are the
backbone of AI applications, ensuring
that vast amounts of data can be
stored, accessed, and processed
efficiently and reliably. Innovations
in these areas directly impact our
Computer-on-Modules' performance,
scalability, and cost-effectiveness."
5
5 Experts on Addressing the Hidden Challenges of Embedding Edge AI into End Products
